This paper presents a new analysis of verbal enclitics and auxiliary verbs that encode aspect, evidentiality, and modality in Shuhi (a.k.a. Shixing, Xumi, ISO 639-3 sxg), a Tibeto-Burman language of Southwest China. Shuhi is underdescribed, and the few previous studies of its verbal enclitics and auxiliary verbs differ considerably in their analysis. To give one example, the auxiliary verb dʑɑŋ is analyzed as an indirect evidential marker (Huang and Renzeng 1991), a direct evidential marker (Sun et al. 2014), and a durative marker (Chirkova 2009, 2017). The analysis in this paper is based on a corpus of 80 interlinear texts that include, among others, translations from Mandarin Chinese, elicitation with non-verbal stimuli, personal narratives, folk stories, procedural texts, conversations, as collected with over twenty speakers from the three main sub-varieties of the language of the upper, middle, and lower reaches of Shuiluo River (the corpus is available at elararchive.org). In addition to providing an updated analysis of verbal enclitics and auxiliary verbs that encode aspect, evidentiality, and modality, the scope of the corpus (the range of genres and the number and provenance of speakers) facilitates addressing two likely causes for the notable differences in previous descriptions of Shuhi. Specifically: (i) differences in data sets used for analysis: making greater use of elicitation from Mandarin Chinese in Sun (1983), Sun et al. (2014), and Huang & Renzeng (1991), as opposed to greater reliance on spontaneous discourse in Chirkova (2009, 2017) (ii) differences between the sub-varieties of Shuhi, possibly reflecting contact influence from different languages: Sun (1983) and Sun et al. (2014) focusing primarily on the sub-varieties of the lower and middle reaches of Shuiluo River, which are in contact with the Pumi and Mosuo languages; and Huang & Renzeng (1991) and Chirkova (2009, 2017) focusing on the sub-variety of the upper reaches of Shuiluo River, which is in contact with Kami Tibetan. The present paper presents a system of aspect, evidentiality, and modality in Shuhi that is more elaborate than previously recognized in terms of the number of grammatical elements and their interrelationships. The system is furthermore argued to be (i) shared by all sub-varieties of Shuhi, and (ii) both similar to and different from the corresponding systems in its contact languages. In relation to past situations, Shuhi exhibits a complex interplay between the lexical semantics of the verb (controllable and non-controllable, internal and observable state verbs), aspect, and the egophoricity-evidentiality parameter, as has also been commonly noticed in the neighboring languages (e.g. Sun 1993; Lidz 2010; Tournadre 2011; Tournadre & LaPolla 2014; Daudey 2014). An example is provided in (1), where the controllable verb ‘to sit’ is followed by the direct (visual) evidential marker; the perfective forms of the verbs ‘come’ and ‘sit’ are derived by adding (telicity-inducing) directional prefixes to the verb stem: (1) zɑ̃³⁵ dʑĩ tjɐ⁵⁵ kʰu⁵⁵-ljɐ⁵⁵ se sɛ³³ɹɛ⁵⁵ lɑ mjɐ³³-zũ⁵⁵ tɕyɐ son one there inward-come CONJ chair LOC downward-sit VS [The speaker watches a video clip describing what she sees:] ‘A man has come in and sat on the chair.’ (Staged events, 079ET_chairsit) In relation to non-past situations, on the other hand, Shuhi has a more elaborate system of epistemic modality than its contact languages, as well as a distinct grammaticalized category of irrealis (counterfactual conditionals, optatives, recommendations), see example (2). (2) ŋɐ⁵⁵ ʁu³³-lɑ⁵⁵ ɦɔ̃⁵⁵ dʒɑ̃³⁵ ʁo ʑy dʑɑ³³hɔ̃⁵⁵ dʑĩ jɛ⁵⁵=ɹɛ̃³³=hĩ⁵⁵ 1SG on-LOC silver have IRR COND house one buy=PRSP=NMLZ.AGT uɐ̃³³=i⁵⁵ COP=NMLZ ‘If I had money, I would be able to buy a house.’ The presentation details the morphological structure of the verbal paradigm, and the meanings of individual markers, their interplay, and their correlations with directional prefixes. In view of the notable differences in previous descriptions of Shuhi, the presentation concludes with reflections on available methodologies of data collection and analysis, and strategies that could be most adapted for recovering complex linguistic structures at the intersection of evidentiality system with other semantic and pragmatic systems in underdescribed languages.